Jose reviewed an interesting journal article: "Path Data in Marketing". You can read it at that link, at the Web Analytics Assocations' Research Committees' Peer Review Journal's Project.
And, it got me into thinking more about path analysis.
To a web analyst, traditionally, a path analysis is examining a sequence of pages that were viewed. Back in the nineties we used to call such analysis 'threading': and we always chose to examine pages and the sequence.
Threading was computationally expensive during the nineties, when volume was low, and it continues to be very computationally heavy for vendors to this day - even with improved algorithms: the volume hurts.
How much does Page Path Analysis (Maddeningly: even when we use the word "click path analysis", we don't really mean 'click', we mean 'pageload'.) really tell us these days anyway?
On highly interactive density sites: not much. One of the most common behaviors on websites is the single page visit. On this point, we treat all 'bounces' like a fail, when in reality, I think there needs to be a differentiation between a 'bounce rate', and a 'reject rate'.
It's entirely possible for a user to visit a website, engage with the copy, and leave to research more (a single page visit) . To the web analyst, that single visit is as a fail. Granted, a more seasoned web analyst would have a filter to differentiate return visitor conversion from first visitor conversion. If they could get that filter or that dimension cut. (Not to trivialize it: but it IS hard in some organizations).
Alright, so this, naturally, goes back to what's wrong with the 'time on site' metric too. A single page visit, with 30 minutes worth of active engagement, would still be treated by most web analytics tools as being a bounce. If you believe what you read in some books, you're told that a bounce is a fail all the time. It is not a fail all the time.
If you're a blogger, for instance, you want to know that people came, they read, they screwed off. I'm not hurt if people don't want to check out my other posts. My other posts are not relevant to everybodys' interests. I'm successful if people are still skimming by the last paragraph. At present, I can't see that. Some bloggers want you to click on their ads, because that's how they get paid, and so, that's a different form of path, isn't?
When we only hit the web analytics server with a pageview, then all we're measuring is pages. And there's more to the path than pages.
There are specific actions that a user does that could be very indicative of interest and/or success. For instance, if a user lands on a page, and leaves before a 7 second load flag has been fired by a piece of Javascript, then we can treat that as a rejection. If a user lands on a page and scrolls slowly, the odds are good that they're reading. If the mouse activity is heavily highlighting, or click on images - then that's indicative of engagement. (For instance, we've trained users over the years to expect that when an image is clicked upon, it becomes bigger - for more detail). If a user begins filling out a form, or engages with a piece of JQUERY or an adverarea - then that's another form of success. If the user has scrolled down as far as they can scroll: that too is success to some people on some pages.
In fact, the very idea of the 'microgoal', or 'goal of a page' is something that should come on back. The notion that some pages are routers, some are converters, some are completers, some are decision-support, and others are branding - I think is important. You WANT people to spend a lot of time with branding pages - but you don't want people leaving at a router page. Sometimes you want people to take a very specific action on that specific visit - and you want them to take another action at a different time.
The broader challenge of path analysis is journey analysis. Stringing individual paths together to understand how multiple visits, over time, add up to a buying or a desired action. This, of course, is totally hard because people frequently begin and end their journeys on different devices at different times. That is not to say that we can't measure the broader patterns and do the legwork to unify each part of the path.
That said, it would be far easier to take on journey analysis if we had a more robust path analysis available.
Tuesday, July 28, 2009
Thursday, July 23, 2009
What meme-tracking can teach us about text vizualization
"Meme-tracking and the Dynamics of the News Cycle" is an excellent paper put out by Leskovec, Backstrom and Kleinberg.
In it, they track meme's against the news cycle. The empirical findings of their study, which focuses on Palin, is really cool. How they chose to vizualize what was going on that's quite new.
Traditionally, we tend to graph social networks using graph theory: each person is a node, and you draw links. Sometimes we color in the nodes and represent the strength between nodes by thickness of the lines. This kind of social network vizualization is something very cool, but the type of math that's required to derive a real business strategy out of it is not. People have a hard enough time deciding what to do to reduce "bounce rate", little though "eigenvector centricity". But, there's competitive advantage is making something hard to do - easy.
All social networks can be expressed as a mathematical matrix - where if two people know each other, we populate a '1' at the intersection. You can also populate it with an interger or a float, variously to represent the strength or the nature of the relationship between two people.
What I like about what was done here is that the authors populated each node with a pointer to more information. There's something about this transmission of text throughout a social network - how it evolves, twists, mutates and spins - that really is something really quite special. Hopefully the graph below is clickable and you see it better.
Think of it from a marketer's point of view.
Your message, even if you do a good job of seeding it to the right people, will not be spread in quite the same way that you want it to be spread. Expect it to shift and to evolve. Sometimes you'll like the evolution - because it'll drive more people to your site and sales. Sometimes you won't - because people won't like you or what they hear.
For certain companies, at certain levels of social media spend - it's a worthy part of their social analytics programme. It's a lot nicer to see the evolution of what people are saying as a message is getting transmitted in conversation than to look at a tag cloud that doesn't really tell you all that much. There's something more human, and something vastly more conversational about vizualizing conversations over long periods of time in this way.
I applaud Leskovec, Backstrom and Kleinberg. Great work. Thank you.
In it, they track meme's against the news cycle. The empirical findings of their study, which focuses on Palin, is really cool. How they chose to vizualize what was going on that's quite new.
Traditionally, we tend to graph social networks using graph theory: each person is a node, and you draw links. Sometimes we color in the nodes and represent the strength between nodes by thickness of the lines. This kind of social network vizualization is something very cool, but the type of math that's required to derive a real business strategy out of it is not. People have a hard enough time deciding what to do to reduce "bounce rate", little though "eigenvector centricity". But, there's competitive advantage is making something hard to do - easy.
All social networks can be expressed as a mathematical matrix - where if two people know each other, we populate a '1' at the intersection. You can also populate it with an interger or a float, variously to represent the strength or the nature of the relationship between two people.
What I like about what was done here is that the authors populated each node with a pointer to more information. There's something about this transmission of text throughout a social network - how it evolves, twists, mutates and spins - that really is something really quite special. Hopefully the graph below is clickable and you see it better.
Think of it from a marketer's point of view.
Your message, even if you do a good job of seeding it to the right people, will not be spread in quite the same way that you want it to be spread. Expect it to shift and to evolve. Sometimes you'll like the evolution - because it'll drive more people to your site and sales. Sometimes you won't - because people won't like you or what they hear.
For certain companies, at certain levels of social media spend - it's a worthy part of their social analytics programme. It's a lot nicer to see the evolution of what people are saying as a message is getting transmitted in conversation than to look at a tag cloud that doesn't really tell you all that much. There's something more human, and something vastly more conversational about vizualizing conversations over long periods of time in this way.
I applaud Leskovec, Backstrom and Kleinberg. Great work. Thank you.
Friday, July 17, 2009
Marketing to an older generation
Paco Underhill had an interesting interview on the NewsHour with Jim Lehrer that was re-aired this week.
It struck a cord. I'm still thinking about it 72 hours after the fact. You can read the transcript here.
While there is a lot of good content in there, it was this part that really caught me [my emphasis]:
Internet Marketing really is dominated by twenty and thirty somethings. It's an intensely young industry. There are exceptions of course, and I really do respect the older people in our field. I like to think that they make us wiser and that we make them younger. To Paco's point: he's right.
We've been building a lot of one-size-fits-all experiences for 15 years now. I'd argue that many experiences are built for the market segments / personas that we think are the customers.
I've been lucky though. I've worked on over fifteen projects aimed at those 55 and older. I know of some people who have gone their entire career without ever marketing to those people. For the majority of us though, he's right.
I think it's time to use data to make experiences better for the older generations. For instance, if a user tells us through their behavior or through their preferences that they don't like drop down menus, then there shouldn't be drop down menus. (And that's just the tip of the iceberg of what is doable! See the posts on Website Morphing to get an idea.).
Web Analysts have a tremendous opportunity here over the next five years to really contribute to the solution.
Either that - or start applying for jobs at wine and pasta concerns.
:P
It struck a cord. I'm still thinking about it 72 hours after the fact. You can read the transcript here.
While there is a lot of good content in there, it was this part that really caught me [my emphasis]:
Marketing to a younger generation
PAUL SOLMAN: Well, let's say there are some people in our audience who would like to re-inflate the bubble. How do you get consumers to start buying again at this point?PACO UNDERHILL: Nobody's going to go back to the old ways. And what we're seeing here is a time in which our retail world is probably going to contract.
It is going to contract, and that's because we are over-stored, meaning that most retail entities would be eminently healthier if they were smaller. Sixty percent of discretionary income in North America is held in the hands of people who are 55 and over.
PAUL SOLMAN: And we don't need stuff?
PACO UNDERHILL: Paul, you and I could live the rest of our lives on fruit, vegetables, pasta, wine, olive oil, and yearly doses of socks and underwear.
I think the other thing that is interesting is that our basic marketing engines are in the hands of people who are 30-something. And they like selling to themselves, and they like selling to a younger generation. They're not that comfortable selling to gray, bearded, bald, paunchy research wonks like you and I.
Internet Marketing really is dominated by twenty and thirty somethings. It's an intensely young industry. There are exceptions of course, and I really do respect the older people in our field. I like to think that they make us wiser and that we make them younger. To Paco's point: he's right.
We've been building a lot of one-size-fits-all experiences for 15 years now. I'd argue that many experiences are built for the market segments / personas that we think are the customers.
I've been lucky though. I've worked on over fifteen projects aimed at those 55 and older. I know of some people who have gone their entire career without ever marketing to those people. For the majority of us though, he's right.
I think it's time to use data to make experiences better for the older generations. For instance, if a user tells us through their behavior or through their preferences that they don't like drop down menus, then there shouldn't be drop down menus. (And that's just the tip of the iceberg of what is doable! See the posts on Website Morphing to get an idea.).
Web Analysts have a tremendous opportunity here over the next five years to really contribute to the solution.
Either that - or start applying for jobs at wine and pasta concerns.
:P
Monday, July 13, 2009
Social Media Analytics: the hidden cost of 'free'
As with any early industry, there are quite a few 'free' measurement applications out there. Social Media is no exception.
Many of them are quite good at doing one or two things very well. For instance, Twitalyzer is very good at measuring influence, and in particular, the first derivative of influence. The advanced search functions on Google are very good at tracking, at least at a monthly cadence, the number of mentions and backlinks. Very useful. And they're FREE*!
There's a large component of social media analytics that can be done with Google Analytics for 'free' too.
Of course, 'free' has a hidden cost. In some instances, unless you're opting out and getting mutual NDA's, you're giving up some privacy. Of course, privacy has little value to many people anyway - it might as well be free. (You'd be amazed how many people will trade their SSN for as little as a five dollar gift certificate to burger king).
There's also the hidden cost of aggregation and interpretation. It all depends, of course, on what you're actually going to do with it.
In certain situations, when you and you alone are the decider:
you don't need to put a lot of effort into collecting and aggregating the information into a single neat place so that a team can see what's going on.
In most instances though, there are many people who want to see what's going on, and they don't want to log into 9 different tools just to see what is being said about them online.
Then there's the interpretation and processing of information. Try coding 100 comments into 'positive', 'neutral' and 'negative' by hand to see what I'm talking about....
All told, 'free' really isn't free.
That's not to say "go out and experiment". By all means, please, go out and experiment. Check them out. They're great tools. They're great interfaces.
When you're setting expectations on a social media strategy, just don't assume that you suppress the costs of measurement and optimization down to zero or to 'free'. It's just not true, and you're setting yourself up for some real misery down the road.
Not to kill your contact high. It's still exciting.
Many of them are quite good at doing one or two things very well. For instance, Twitalyzer is very good at measuring influence, and in particular, the first derivative of influence. The advanced search functions on Google are very good at tracking, at least at a monthly cadence, the number of mentions and backlinks. Very useful. And they're FREE*!
There's a large component of social media analytics that can be done with Google Analytics for 'free' too.
Of course, 'free' has a hidden cost. In some instances, unless you're opting out and getting mutual NDA's, you're giving up some privacy. Of course, privacy has little value to many people anyway - it might as well be free. (You'd be amazed how many people will trade their SSN for as little as a five dollar gift certificate to burger king).
There's also the hidden cost of aggregation and interpretation. It all depends, of course, on what you're actually going to do with it.
In certain situations, when you and you alone are the decider:
you don't need to put a lot of effort into collecting and aggregating the information into a single neat place so that a team can see what's going on.In most instances though, there are many people who want to see what's going on, and they don't want to log into 9 different tools just to see what is being said about them online.
Then there's the interpretation and processing of information. Try coding 100 comments into 'positive', 'neutral' and 'negative' by hand to see what I'm talking about....
All told, 'free' really isn't free.
That's not to say "go out and experiment". By all means, please, go out and experiment. Check them out. They're great tools. They're great interfaces.
When you're setting expectations on a social media strategy, just don't assume that you suppress the costs of measurement and optimization down to zero or to 'free'. It's just not true, and you're setting yourself up for some real misery down the road.
Not to kill your contact high. It's still exciting.
Thursday, July 9, 2009
The effect of the Internet on Prices
One of the more neat aspects of the Internet is its impact on prices.
Before the Internet, researching the best price for something was relatively hard. Or, I imagine it to have been hard. The price you got for a consumable, like a car, a house, an airline ticket or hotel room largely depended on who you knew or how many people you called and asked.
One of the impacts of the Internet has been relative deflation in prices as a result of the ability of customers to compare prices easily. This decrease in the cost of becoming un-ignorant has eaten directly into the margin. I don't think I can argue that the barriers of entry have been significantly lowered as a result of the Internet: a hotel still requires capital to build rooms, and an airline still requires airplanes and people to fly. But I can see that certain barriers have come down. It's possible for certain companies to compete entirely on price and cost-efficiency models. Customer service is increasingly becoming a relevant differentiator too. Refreshingly.
You also have a whole bunch of aggregator companies these days too who derive their income by making the cost of research to consumer (in terms of time, effort, and ease) come down as well. In way, this represents a refreshing triumph of ingenuity and innovation: how revenue management analytics, web analytics, creative, information architecture, and strategic IT can come together to create revenue streams that didn't exist before.
We could see still more deflation on the margins. I'm more interested to see how companies discover new values that "consumers didn't know they would want to pay for".
Before the Internet, researching the best price for something was relatively hard. Or, I imagine it to have been hard. The price you got for a consumable, like a car, a house, an airline ticket or hotel room largely depended on who you knew or how many people you called and asked.
One of the impacts of the Internet has been relative deflation in prices as a result of the ability of customers to compare prices easily. This decrease in the cost of becoming un-ignorant has eaten directly into the margin. I don't think I can argue that the barriers of entry have been significantly lowered as a result of the Internet: a hotel still requires capital to build rooms, and an airline still requires airplanes and people to fly. But I can see that certain barriers have come down. It's possible for certain companies to compete entirely on price and cost-efficiency models. Customer service is increasingly becoming a relevant differentiator too. Refreshingly.
You also have a whole bunch of aggregator companies these days too who derive their income by making the cost of research to consumer (in terms of time, effort, and ease) come down as well. In way, this represents a refreshing triumph of ingenuity and innovation: how revenue management analytics, web analytics, creative, information architecture, and strategic IT can come together to create revenue streams that didn't exist before.
We could see still more deflation on the margins. I'm more interested to see how companies discover new values that "consumers didn't know they would want to pay for".
Saturday, July 4, 2009
Is "Private Browsing" really "Private"?
Web analysts I speak with are expecting a crash in observed traffic with the launch of Firefox 3.5 and the newer features within IE.
I question if it's really all that private at all.
Some commentators, like Preston Gralla, call private browsing the "porn mode". Gralla goes onto write:
A record, though not linked to anything personally identifiable on its own, is kept on the porn sites' log files (and, if the site requires javascript, I don't really see how private browsing can keep all the web analytics script from firing if it's embedded right into the experience).
The concern of many web analysts is that just as a portion of the population are habitual cookie deleters, there are going to be people who turn on private browsing and leave it on.
What do cookies really do for us in terms of measurability?
Well, assuming that nobody ever deleted their cookies, it would have let me know with a high degree of accuracy, which browsers on a specific computer were returning to the website at a specific recency and frequency. Such information is important to a blogger because you can judge, roughly, just how much audience you're retaining - and make decisions on frequency and content. Of course, cookies don't measure people. Multiple people use the same computer, and a single person uses multiple computers over the run of day. It's a proxy measure of return visitors though, and it's still useful to an analyst so long as they know the real definition.
Another big use is for campaign attribution. Certain websites get paid on performance - meaning that they only get money if you click through on an ad and buy something - typically within 30 and 45 days. Without a cookie - they don't get paid. It makes me wonder if certain websites that get paid for performance just might as well throw up the rule "Want to visit this website? Turn off Private Browswing" as a way to deal with it. I know many developers have been having similar fantasies about denying access to those who continue to use IE 6. In the end, I think that it's the pay-for-performance business model that might end up suffering the most. (And I can't imagine what kind of damage this is going to do to for the porn affiliate programs).
A third big use is for analysts is personalization and making your site experience better: like remembering your username or associating a certain behavior pattern with being a good customer and getting perks that encourage more intense behavior.
Here's a big takeaway:
Your browsing behavior can be recorded by software installed on the client side. The search engines are most certainly recording which queries you're entering. It most certainly is being recorded on the ISP side. Who the hell knows who else is looking at that data from the ISP side after the spliter. Then it's being recorded, at some level, by third party ad servers and analytics companies. And, at the source of where the information is being housed, by server logs. If I was a rational policy analyst or a citizen concerned about surveillance, I'd be far more worried about those touchpoints where the information is personally identifiable.
Let me explain.
The web analyst sitting in front of the ad server report, the server logs, or the nicely formatted web analytics data has a hard enough time interpreting that data at the aggregate. 99.999% of the time, we have no idea of the identity of the person we're looking at. In that 0.001% instance, it's usually a filter I've set up specifically to track my own behavior. (What sort of analyst would I be if I didn't understand how my own behavior is reflected on the tracking software that I use?)
The ISP's, and the people behind the splitter: they know the billing addresses and are far closer to your identity. The person in your household that has installed some software one of your computers: they know exactly who you are and what you're doing.
Privacy on the Internet is such a delusion.
Private browsing really isn't private.
I question if it's really all that private at all.
Some commentators, like Preston Gralla, call private browsing the "porn mode". Gralla goes onto write:
When you browse the Web using it, nothing about the session is stored -- no history, no cookies, no temp files, no forms information, no search information, nothing that can show where you've browsed or what you've done. To turn the Las Vegas tag line on its ear: What happens in Firefox doesn't stay in Firefox.Well, alright, nothing in-the-browser is really stored. Of course, some memory of the porn run is recorded on the hard-drive and by any sort of spyware that a spouse, parent, partner, or roommate has installed on the computer.
A record, though not linked to anything personally identifiable on its own, is kept on the porn sites' log files (and, if the site requires javascript, I don't really see how private browsing can keep all the web analytics script from firing if it's embedded right into the experience).
The concern of many web analysts is that just as a portion of the population are habitual cookie deleters, there are going to be people who turn on private browsing and leave it on.
What do cookies really do for us in terms of measurability?
Well, assuming that nobody ever deleted their cookies, it would have let me know with a high degree of accuracy, which browsers on a specific computer were returning to the website at a specific recency and frequency. Such information is important to a blogger because you can judge, roughly, just how much audience you're retaining - and make decisions on frequency and content. Of course, cookies don't measure people. Multiple people use the same computer, and a single person uses multiple computers over the run of day. It's a proxy measure of return visitors though, and it's still useful to an analyst so long as they know the real definition.
Another big use is for campaign attribution. Certain websites get paid on performance - meaning that they only get money if you click through on an ad and buy something - typically within 30 and 45 days. Without a cookie - they don't get paid. It makes me wonder if certain websites that get paid for performance just might as well throw up the rule "Want to visit this website? Turn off Private Browswing" as a way to deal with it. I know many developers have been having similar fantasies about denying access to those who continue to use IE 6. In the end, I think that it's the pay-for-performance business model that might end up suffering the most. (And I can't imagine what kind of damage this is going to do to for the porn affiliate programs).
A third big use is for analysts is personalization and making your site experience better: like remembering your username or associating a certain behavior pattern with being a good customer and getting perks that encourage more intense behavior.
Here's a big takeaway:
Your browsing behavior can be recorded by software installed on the client side. The search engines are most certainly recording which queries you're entering. It most certainly is being recorded on the ISP side. Who the hell knows who else is looking at that data from the ISP side after the spliter. Then it's being recorded, at some level, by third party ad servers and analytics companies. And, at the source of where the information is being housed, by server logs. If I was a rational policy analyst or a citizen concerned about surveillance, I'd be far more worried about those touchpoints where the information is personally identifiable.
Let me explain.
The web analyst sitting in front of the ad server report, the server logs, or the nicely formatted web analytics data has a hard enough time interpreting that data at the aggregate. 99.999% of the time, we have no idea of the identity of the person we're looking at. In that 0.001% instance, it's usually a filter I've set up specifically to track my own behavior. (What sort of analyst would I be if I didn't understand how my own behavior is reflected on the tracking software that I use?)
The ISP's, and the people behind the splitter: they know the billing addresses and are far closer to your identity. The person in your household that has installed some software one of your computers: they know exactly who you are and what you're doing.
Privacy on the Internet is such a delusion.
Private browsing really isn't private.
Subscribe to:
Posts (Atom)